Add Qwen3.6 validation and benchmark coverage

tools/bench/config.py

@@ -25,10 +25,7 @@ class SystemEndpoint:
     model_id: str                  # what to put in the request body's "model" field
     api_key: str | None = None     # llama-server doesn't need one; xserv ignores it
     # Extra fields merged into every request body for this system. Used to keep
-    # the two engines in the SAME generation mode — xserv hardcodes Qwen3
-    # thinking OFF (empty <think></think> in its prompt builder), so we disable
-    # thinking on llama-server via chat_template_kwargs to match. Both engines
-    # ignore unknown fields, so this is safe.
+    # both engines in the same chat-template generation mode.
     extra_body: dict | None = None
     # Process supervision is optional — if base_url is already serving, we skip launch.
     launch_cmd: list[str] | None = None
@@ -49,7 +46,12 @@ class BenchConfig:
     quality_max_tokens_aime: int = 16384
     quality_max_tokens_gsm8k: int = 2048
     quality_limit: int | None = None   # subsample for smoke tests; None = all
+    quality_seed: int | None = None
     quality_temperature: float = 0.0
+    quality_top_k: int = 0
+    quality_top_p: float = 1.0
+    quality_presence_penalty: float = 0.0
+    quality_repetition_penalty: float = 1.0
     request_timeout_s: float = 1800.0